Anyone here have a rough idea about what volume of cars you need to have coming off a production line for it to be efficient to mass produce? My google-fou has failed me.
Just noodling around with the idea that we’ll know that AV services are going to be in a position to possibly seriously scale when they approach about 3-4x the minimum size of a production run. All of these AV TaaS efforts are going to be economically disadvantaged against Uber/Lyft/Taxi until they reach the point where they’re no longer using “bespoke” modified cars. It’s just really expensive to take a car that’s not suitable for AV TaaS and individually modify it with the extra communication and sensor devices necessary to be part of a TaaS fleet.
The problem, though, is that even the biggest AV TaaS fleet (Waymo’s) is so small that it can’t yet be economical for them to fire up a mass production line for their vehicles. Waymo’s only got about 2,500 vehicles - even if they only had 3-4 years of useful life, they can’t need as many as a thousand new vehicles per year. Which probably isn’t enough to justify running an actual production line. And if they produced more, where would they put them?
If the minimum run to make a production line efficient is 10K (to pick a number out of thin air), though, then once a TaaS service started to approach 30-40K vehicles in their fleet, it might make sense to fire up that production line. Because then you’d have enough “spots” in your service for those cars to land in.
Mine says: “ Economies of Scale: To be financially viable for major automakers, a single platform often requires production of 250,000 to 1,000,000+ vehicles per year”
Of course that’s for a “normal” car. For one which is income generating it might be less, given that you would have (some) more expense on the front end to be recouped with the operation of the business. That said, (and as we have both argued) the lifespan of a taxi is likely to be less than that of a family car, but then again you don’t have to sell them wholesale to a dealer who marks them up to retail for customers.
Then again, there’s the whole back end of the business to worry about: parking lots, charging apparatus, cleaning, billing, and all the rest - so there are a lot of ifs and buts and candies and nuts to any analysis of this business, which I don’t think any of us are truly qualified to address.
I recall this discussion during Fords discussion of the Pinto explosions when rear ended. The $5 bracket that would have protected the fuel tank would cost millions because it was needed on every vehicle. And they needed to make millions of them to fill distribution channels—while still figuring out if people would buy it.
The bracket would have cost $5 per vehicle. Lee Iacocca had mandated the engineers bring the design in at “under 2,000 pounds and under $2,000 dollars”. So they left off the bracket, even though their crash tests showed it made the car vulnerable to gas tank explosions during preliminary crash tests - and yes, before production was “frozen.”
It was at that point they did the “cost/benefit” which claimed paying damages after the fact was cheaper, in aggregate, than putting bracket on the back to stop the rear bumper fro puncturing the tank in a rear end collision.
I’ve dug in a little more, and it appears that for a typical modern automotive line (about a billion dollars in investment), manufacturers look to have a minimum run of about 50K per year for 5 years to amortize that investment across in order for that to be considered a viable outcome. That’s probably not the minimum, of course - and modern production lines are flexible to be able to switch between different models (especially if they use the same basic platform).
However, in looking at AV’s, I think it’s probably a safe bet that for a truly bespoke purpose built robotaxi, a company probably needs somewhere a bit north of 10K per year for that to be a materially better option than simply modifying other cars. With a useful life of at least several years, that means the “threshold” for one of these robotaxi services to be able to save money/reap economies of scale by switching to mass-produced cars would be a fleet size of 30-50K.
So it will be interesting to watch Waymo and Tesla’s production volumes at Mesa and of the Cybercab, respectively. They’re probably looking at very high per-vehicle costs until they get to a pretty big fleet size, which neither of them are anywhere near.
Reportedly, the self-driving tech leader is looking to purchase 50,000 Hyundai IONIQ 5 electric cars in the next few years, at a cost of about $2.5 billion.
The IONIQ 5 EVs reportedly would not come equipped with the various lidar, radar, and other sensors Waymo uses. The cost of the vehicles would reportedly cost over $50,000 each initially, before the sensory suite is added. Though, I don’t see how that makes sense. No version of the IONIQ 5 costs close to $50,000 even for normal retail consumers. So, at a cost of $50,000+, Hyundai must be adding something — quite a lot — to the vehicles. Perhaps the original reporting is wrong and the vehicles would come with the various sensors included.
As far as we know, the the contract has not been confirmed. However, if details have leaked out, one would think that a final decision and an announcement are coming.
There’s also the IONIQ 5 N, which is their super performance vehicle at $67K, but that seems unlikely.
10K per year still seems small enough that it makes sense to buy “stock” cars and then convert them to AV, rather than for Waymo to have their own mass produced AV car that comes off the line AV ready.
OK, but we all know MSRP is a fiction price. Invoice price is generally 5% below MSRP, and then there are the holdbacks, incentives, floor planning or other words used to reduce the “actual factory price” below that, generally somewhere between 2-5%. So it appears that the real factory price is going to be somewhere south of 10% below MSRP - and given the size of the order you’d expect some serious clawbacks as well.
But if it’s 10%-12%, that makes a $50,000 car (on your chart) actually cost around $43-$45,000. Still not enough to account for all the extra doodads that Waymo has to add, but it’s possible on a single order this large that much of the wiring, harnesses, etc. could be done “on the line”, bringing the Waymo additional cost of outfitting down significantly.
Sure - but it’s not hard to see how that still kind of works out for “ordinary” IONIQ 5’s. The article quoted $2.5 billion for the 50K cars over five years. Maybe in year 1, the cost to Waymo might be $45K. Year 2, $47K - year 3, $49K, year 4 $52K and year 5 $55K - with more volume in the later years than the earliest ones. You kind of end up in the ballpark of wholesale prices to a bulk purchaser once you take that into account. No doubt Hyundai is doing some stuff to accommodate their end user, but we can’t look at the purchase price and assume they’re putting too much extra into the car in their plants (as opposed to Waymo’s Mesa facility).
Excepting the actual cameras & lidar & such (big exception, I know), the huge cost to them is taking apart the entirety of the car and running wires and cables all over the place and then reassembling it. If you could just get some extra cabling done “on the line”, you would probably eliminate half the upgrade cost.
Absolutely. I think that’s certainly going on, and I don’t think estimates of Waymo’s costs per vehicle based on info from a few years ago (where they had to eat those teardown/rebuild costs) are relevant to their new vehicles.
Tesla has multiple assembly lines at his plants. He could control overhead costs by campaigning robotaxi production. Make robotaxis for x months and then switch to other models for the rest of the time.
This is absolutely true. For the second half of my career (as an engineer) I worked in an industry that was related to vehicles and subsystem manufacturing. Often we needed full vehicle wiring harnesses for testing (along with various test equipment that simulates signals at different parts of the harness). Once we received a full harness from Toyota and we spread it across our large conference room table (probably 30 or 40 ft long) and it was huge and very complex. I thought I had a photo, but I can’t find it right now (why can’t AI properly search my photos for “conference room table with wiring harness on it”?)
The reason it would lower the cost significantly is because the total time would go way down. Getting into vehicle crevices and cutting into existing wiring is VERY time consuming. I’ve done it myself and it is very annoying, and you have to do it all correctly the first time because tracking down errors is really tough.